Environ. Sci. Technol. 1989, 23, 75-83
Measurements of the Gas/Particle Distributions of Atmospheric Organic Compounds Mary P. Ligockit and James F. Pankow* Department of Environmental Science and Engineering, Oregon Graduate Center, 19600 N.W.Von Neumann Drive, Beaverton, Oregon 97006
The gas/particle distributions of polycyclic aromatic hydrocarbons (PAHs) and other organic compounds were measured at ground level in Portland, OR, and at the Oregon coast. Measured particulate-phase concentrations of most PAHs and oxo-PAHs were virtually the same whether glass fiber or Teflon membrane filters were used. Backup glass fiber filter concentrations were measured and found to be quite variable, averaging 30-70% of the primary filter values for alkanes, but less than 30% for all PAHs. At the coastal site, due to the lower concentrations of particulate matter there, the gas/particle distributions of PAHs were shifted toward the vapor phase relative to what was observed in Portland. At both sites, for the PAHs of low to intermediate volatility, the gas/particle distributions correlated with the supercooled liquid vapor pressures. However, acenaphthylene and fluorene showed more association with the particulate phase than predicted by those correlations. Possible explanations for the deviations include the presence of bound PAHs within particles and the presence of high energy adsorption sites on the surfaces of the particles. Introduction Atmospheric organic compounds are found in both the gas and particulate phases. Pankow (1)has recently reviewed the theories available for describing this distribution process. The manner in which a given compound partitions will influence that compound's atmospheric removal mechanisms and lifetime, as well as its health effects due to inhalation. The extent of association with particulate matter will depend upon the compound's vapor pressure, the amount and type of particulate matter present, and the temperature. The first equation developed to describe this partitioning process was derived by Junge (2) based on a linear Langmuir isotherm
4 = cp/(cg + cp) = c 8 / ( p 0 + c8)
(1)
where 4 is the fraction of a compound that is associated with particles, cg and cp are the gas- and particulatephase-associated atmospheric concentrations (ng/m3), respectively,p" is the vapor pressure of the pure compound (Torr), and 8 is the total suspended particulate surface area concentration (cm2/cm3). Eq 1 gives CK 1-4 log = log -= CP
4
1 log + log p" c8
(2)
Junge (2) used c = 0.13 to illustrate the dependence of 4 on pa and 8. Equation 1is useful for demonstrating the general trend of increasing partitioning to particles with decreasing p". However, the assumptions implicit in its derivation should be kept clearly in mind. First, it only describes adsorption that is physical, nonspecific, and a t equilibrium, with Present address: Environmental Quality Laboratory, California Institute of Technology, Pasadena, CA 91125. 0013-936X/89/0923-0075$01 SO10
100% of any particulate-associated material assumed to be available for exchange (within the time frame of interest) with the gas phase. Second, one cannot expect that the value of c will be constant from compound class to compound class, or even necessarily within a given compound class. Indeed, as shown by Pankow (1)for physical adsorption, c will be a function of temperature T (K)and will only be constant for a given group of compounds when there is a compound-to-compound constancy in (1) the difference between the entropy of desorption from the specific particulate matter of interest and the entropy of vaporization of the pure compound and (2) the difference between the enthalpy of desorption from the particulate matter and the enthalpy of vaporization of the pure compound. The measurement of cg/cpdistributions is often accomplished by using a filter followed by an adsorbent such as polyurethane foam (PUF) or Tenax. In one of the first studies of this type, Cautreels and Van Cauwenberghe (3) provided measurements for polycyclic aromatic hydrocarbons (PAHs), alkanes, phthalate esters, and fatty acids. Yamasaki et al. (4) used a similar sampling approach and employed a linear Langmuir isotherm to examine the dependence of c /cp distributions of PAHs in Tokyo upon T and the tod suspended particulate matter concentration (TSP). They suggested that 8 might be proportional to TSP and correlated their data using the equation
where A and F are the measured cg and cp values as determined on the adsorbent and filter, respectively, and m and b are compound-dependent constants. If there are no sampling artifacts, then A = cg and F = cy Under these conditions, one may view the expression for the constant K as the equilibrium ratio of A to FITSP, Le., as the equilibrium ratio of the concentration in the gas phase (ng/m3) to that in/on the particulate matter (ng/pg). As 1/T decreases, the gas phase tends to be preferred over the particulate phase. As a result, the slope (m)is negative. The value of b will depend in part on the units selected for TSP. Yamasaki et al. (4) chose units of nanograms per cubic meter, but we will use units of micrograms per cubic meter. As shown by Pankow ( I ) , within a given class of compounds, the value of K may be expected to depend upon p" according to an equation of the type log K = log C + log po (4) where C is a temperature-dependent constant. For the PAHs at 20 "C, log C has been found to be -7.5. Assuming no sampling artifacts, eq 3 and 4 yield log cp/cp = log C/TSP + log p o (5) A comparison of eq 2 and 5 leads to the conclusion that C/TSP = 1/09. There is often some uncertainty as to whether the quantities A and F provide good estimates of cg and cp.
0 1988 Amerlcan Chemical Society
Environ. Sci. Technol., Vol. 23, No. 1, 1989
75
Possible sources of artifacts during sampling include partial volatilization of collected particulate material ( 5 ) , adsorption of gas-phase compounds onto filters (6-a), adsorption onto collected particles (9, l o ) , and chemical reactions with photochemically produced oxidants (11,12). Despite these potential problems, Bidleman and coworkers (9, 10, 13, 14) have corroborated the results of Yamasaki et al. (4) and have been successful in applying eq 3 in parameterizing the dependence of A / F on TSP and 1/T for certain PAHs and organochlorines. A significant observation (10,13)is that although there is some scatter in the log K vs 1/T plots that are currently available for selected PAHs and organochlorines, there is also similarity in the sorption characteristics of the particulate matter of different cities. For example, for fluoranthene and pyrene, the values of m and b obtained in Columbia, SC, (13)are similar to those obtained in Tokyo (4). In terms of eq 5, these results can be interpreted in terms of a rough constancy in the temperature-dependent functionality of C for these compounds ( I ) . In the studies cited above, either glass or quartz fiber filters were used. Recently, some researchers have advocated the use of Telfon membrane filters (TMFs) instead of fiber filters. The Teflon surface, often considered to be less reactive than glass or quartz, has been expected to reduce the amount of "filter-catalyzed" oxidation of PAHs. An early study (15) indicated that the concentration of total PAHs in ambient and exhaust samples was higher when measured on TMFs than when measured on glass fiber filters (GFFs). Subsequent studies have differed in the extent to which this effect was observed for specific PAHs. One study found GFF/TMF concentration ratios for pyrene and benzo[a]pyrene of 0.25 and 0.76, respectively (16). Another study found both of these ratios to be -0.85 and also found no difference in the mutagenicities of the samples collected on the different filter media (17). As part of a continuing study of atmospheric organic compounds (1,18-23),this work was conducted to investigate the gas/particle distribution process with both GFFs and TMFs. The contributions to the measured particulate-phase concentrations made by gas adsorption to the GFFs were determined. Experimental Section The GFFs (102-mm diameter) were obtained from Gelman (Ann Arbor, MI). The TMFs (20 X 25 cm sheets, Zefluor, 2-pm pore size) were obtained from Membrana (Pleasanton, CA). The latter were cut into 102-mm circles. PUF of density 0,022 g/cm3 was obtained in 7.6-cm thick sheets and was cut into 5.1-cm diameter plugs (PUFPs). Tenax-GC and -TA were obtained from Alltech Associates (Deerfield,IL). Solvents were distilled-in-glassgrade (B&J, Muskegon, MI). Perdeuteriat,ed PAHs were obtained from KOR Isotopes (Cambridge, MA) and MSD Isotopes (Los Angeles, CA). GFFs were cleaned prior to sampling by baking at 400 "C for 2 h. TMFs were cleaned by two successive 10-min sonic extractions in 60:40 acetone/hexane, followed by air drying, Filters were transported to and from the field wrapped in prebaked aluminum foil. PUFPs were cleaned by Soxhlet extraction for 24 h in 6040 acetone/hexane and dried under a stream of prepurified nitrogen. They were stored and transported in clean screw-capped glass jars with Teflon cap liners. Teflon tape was wrapped around the threads of each jar to provide an airtight seal. Tenax cartridges were cleaned as described elsewhere (24). Sampling took place in an urban residential section of Portland, OR (20,21),and at a nonurban Oregon coastal site [Ft. Stevens State Park (23)]. In some cases, both 76
Envlron. Sci. Technol., Vol. 23, No. 1, 1989
front and backup filters were used. In every case, a front and a backup PUFP followed the filter(s). A branch of the sampler carried a small fraction of the flow from the filter(s) through a set of Tenax cartridges. The maximum/minimum temperature range during all sampling was generally less than 6 "C. The sampling periods were 5-30 h. The face velocities (30-60 cm/s) were comparable to high-volume face velocities. Six sets of samples were obtained in Portland, OR, in February and April 1984 by using GFFs with backup filters. The range of the mean sampling temperatures for the six sets was 5-9 "C. The overall mean temperature was 8 "C. A portion of the data from these samples has been discussed elsehwere (20,21, 24). Four sets of samples were also collected in Portland with duplicate samplers during February and March, 1985; one sampler was equipped with GFFs and the other with TMFs. The range of the mean sampling temperatures for the four sets of samples was 3-9 "C; the overall mean temperature was 5 "C. The flow rates remained constant throughout each sampling period, but ranged from 100 to 190 L/min over the four events. The flow rates were always within 10 L/min on the two samplers. Pressure drops across the filters were -1 psi for both types of filters. Three sets of samples were obtained with TMFs at the coastal site during April 1985. The range of those mean sampling temperatures was 8-10 "C; the overall mean temperature was 9 "C. After sampling, portions of most of the GFFs were first analyzed for particulate organic carbon, particulate elemental carbon, and total particulate carbon (TPC) (e organic + elemental) with a thermal/optical carbon analyzer (25,26). [Inorganic particulate carbon in Portland is low (26).] The GFFs and TMFs were extracted and analyzed as described previously for GFFs (20,21),except that the acid/neutral separation step was omitted for the 1985samples. Briefly, several deuteriated PAHs in acetone were added directly to the filters as internal standards. The filters were Soxhlet extracted in 25 mL of 1:l acetone/methylene chloride, and the solvent volume was reduced to 2 mL in a miniature Kuderna-Danish concentrator. The extracts were cleaned up on 5 mL of 15% deactivated silica gel and then reconcentrated to 200 pL by N2 blowdown. The extracts were analyzed on an HP 5790A gas chromatograph equipped with a capillary column (30 m long, 0.32 mm i.d., 0.25-pm DB-5 film, J&W Scientific, Folsom, CA) that was interfaced to a Finnigan 4000 mass spectrometer/data system (GC/MS/DS). The PUFPs were analyzed similarly, except that 500 mL of 60:40 acetone/hexane was the initial solvent. The Tenax cartridges were analyzed by thermal desorption and GC/MS/DS as described previously (24). Results and Discussion Filter Blanks. Table I gives the mean blank levels of the alkanes, phthalates, and PAHs on the GFFs and TMFs. The GFF blanks were low for the PAHs in agreement with other work (16,27), but nonnegligible for bis(2-ethylhexyl) phthalate. The blanks were nearly uniformly higher on the TMFs. In addition to physical adsorption, absorption of background contaminants into the Teflon polymer matrix (28, 29) might be a possible cause for the high TMF blank levels. For example, Hult (28) has found that using syringes with TFE Teflon seals to analyze gas samples containing low molecular weight hydrocarbons can lead to prolonged cross-contamination problems. For TMFs, McDow (8) has suggested that the blank levels might be lowered by baking them out at a high temperature, e.g., at 300 "C,which is just below the melting point of Teflon.
Table I. Comparison of per Filter Blank Levels for GFFs and TMFs for Sampling in Portland, OR, during February and April 1985, Including the TMF Blank Data for the April 1985 Sampling at the Coast’
compound eicasane heneicosane docosane tricosane tetracosane pentacosane hexacosane octacosane diethyl phthalate dibutyl phthalate butyl benzyl phthalate bis(2-ethylhexyl) phthalate dioctyl phthalate phenanthrene fluoranthene pyrene chrysene
GFF (n =
TMF (n = 91, ng 91 f 60
121 122 162 f 268 213 f 422 291 i 515 241 f 453 214 f 436 170 f 353 104 f 77 271 f 113 59 f 36 944 f 918 25 f 46 1.1 f 1.3 0.3 f 0.7 0.2 f 0.5 0.6 f 0.8
41, ng 3.5 f 4.4 2.0 4.0 2.9 f 3.5 NDb 11 f 16 ND 6.2 f 12.5 7.2 i 14.3 30 f 14 51 f 43 20 f 20 1876 & 1467 2.6 f 5.2 ND ND
100.0
RIC
*
100.0
L
“Data values are means f l u . *ND, not detected. ~
Front and backup filter sample mass amounts were considered nonzero only when they exceeded the mean blank mass amounts at the 95% confidence level. Because of the high TMF blanks for alkanes and phthalates, these compounds generally did not exhibit ambient levels that passed the signficance test as determined on the TMFs. Normalized blanks were calculated by dividing the blank mass amounts by the corresponding sample volumes. When the significance test was passed, the normalized blank was subtracted from the calculated ambient concentration. Filter Detection Limits. With the GC/MS/DS operated in the scanning mode, the instrument mass detection limits for the PAHs, alkanes, and phthalates were all -0.1 ng. Depending upon the volume of air sampled, this translated into concentration detection limits of 0.034.2 ng/m3. No compound more volatile than acenaphthylene was ever found on the filters. For those volatile compounds that were found, the levels were close to the detection limits. When the April 1985 Portland filter samples were reanalyzed for PAHs with the GC/MS/DS in the multiple ion detection (MID) mode, the resulting enhanced sensitivity produced concentration detection limits of 0.0024.01 ng/m3; the most volatile PAH found on the filters remained acenaphthylene. Adsorption of Organic Glass by GFFs. Since the GFFs used exhibit >99% collection efficiency for all particle sizes (30,31),any material found on a GFF backup filter in a blank-corrected quantity greater than -1% of the front filter amount probably represented sorbed gas. Figure 1 shows typical front, backup, and blank GFF chromatograms obtained during 1984 in Portland. As in Figure 1,the backup filters occasionally exhibited a small hump of compounds that occurred earlier in the chromatogram than the hump in the front filter chromatogram. The determinations of total carbon on the filters generally gave backup filter amounts that were 15% of the front filter values, in good agreement with results of others (7) for quartz fiber and silver membrane filters. In this context, it may also be noted that Fitz et al. (17) observed that GFFs produced an average of 48% more solvent-extractable mass for the same volume of air sampled than did Teflon-impregnated GFFs. They concluded that this was due primarily to filter-mediated gas-to-solid conversion of inorganic gases on the GFFs. (The solvent mix they
-
II
b
100.0-
C RIC-
500 615
1000 1230
1500 1845
2000
2500
2500
3115
SCAN T I M E (mtn)
Figure 1. Typical primary (a), backup (b), and blank (c) glass fiber fitter chromatograms obtained from sampling In Portland, OR. Several of the large peaks in each of the chromatograms are internal standard compounds.
used contained methanol, which is capable of extracting some inorganic salts.) Since total “particulate” organic carbon on filters in Portland is usually -20-30% of TSP, and since the GFF gas adsorption effect observed here corresponded to only 15% of total carbon on the front filters, that conclusion is consistent with our results. All of the target compounds found on the backup filters had vapor pressures in the range 104-104 Torr. No target compounds more volatile than phenanthrene were found on any backup filter, though due to the extremely low particulate-phase concentrations of the most volatile compounds, backup concentrations even as high as 30% of their front filter concentrations would not have been detected. The backup filter values were quite variable. For one sampling event, no target compounds except chrysene were detected. For another event, 16-39% of the front filter amounts were found on the backup filter for PAHs in the 10-8-10-4 Torr vapor pressure range. The other events generally fell between these two extremes. Table I1 gives the average values for the backup filters in GC elution order in terms of the (1) mass amounts, (2) equivalent concentrations, and (3) percentages of the front filter amounts. (It may be noted that phenanthrene has a higher solid vapor pressure than does 9-fluorenone, though it does elute after 9-fluorenone on the GC column used here.) For the methylphenanthrenes and fluoranthene, the mean amounts were 29 and 19%, respectively. For the alkanes and phthalates with vapor pressures in the Environ. Sci. Technol., Vol. 23, No. 1, 1989
77
Table 11. OFF Backup Filter Levels in Terms of Mass Amounts, Equivalent Concentrations (Mass/Sample Volume), and Percentages of Primary Filter Values (Means flu) for Sampling in Portland, OR, during February and April 1984 compound
mass amt, ng
equiv concn, ng/ms
Fprirnary
9-fluorenone phenanthrene methylphenanthrenes 9,lO-anthracenedione fluoranthene pyrene benz[a]anthracene chrysene eicosane heneicosane docosane butyl benzyl phthalate
1.6 f 1.9 1.4 f 3.5 6.6 f 4.3 8.8 f 9.9 14 i 23 12 f 24 3.5 f 6.3 14 8.7 37 f 38 64 i 73 37 f 36 110 f 120
0.02 f 0.02 0.03 f 0.07 0.06 f 0.05 0.08 f 0.12 0.10 f 0.14 0.09 f 0.13 0.06 f 0.12 0.15 f 0.21 0.49 f 0.85 0.89 i 1.63 0.37 f 0.54 0.71 f 0.58
21 i 32 5.5 f 13 29 f 30 9.7 f 13 19 f 27 12 f 18 4.0 f 6.7 9.5 f 8.6 71 f 87 52 f 54 24 f 24 28 i 28
% of
critical range, the adsorption was fairly large, with backup filter amounts for eicosane averaging 71% of the front filter values. Because alkanes were not measurable on the TMFs due to blank problems, their adsorption on GFF backup filters could not be compared to that on TMFs. Losses of adsorbed organic material due to volatilization from collected particles were not measured in this study, though conditions that helped to minimize volatilization included (1)low ( 1psi) pressure drops across the filters and (2) small temperature fluctuations during sampling. Van Vaeck et al. (5) recently concluded that volatilization losses can be very problematic in high-volume sampling for compounds of intermediate volatility. However, they did not consider the effects of gas adsorption. The special sampler that they used to minimize volatilization, and therefore standardize against in their studies, employed fresh filter portions for each of a series of short sampling intervals. Minimizing the (sample volume)/ (filter area) ratio will, however, tend to maximize gas adsorption effects since filter sorption capacities are limited. Indeed, it may be noted that Fitz et al. (17) collected 33% more solvent-extractable mass with four sequential 3-h Teflon-impregnated GFFs than they collected with a single, concurrent, 12-h filter. Also, Cadle et al. (7) have concluded that at high filter loadings, the amount of organic material adsorbed on quartz fiber filters is no longer linearly dependent on the total organic carbon filter loading. If a gas adsorption effect similar in magnitude to that found in the present study occurred during the sampling of Van Vaeck et al. (5), then gas adsorption could have been responsible for a significant portion of the artifact that they ascribed to volatilization. This is a definite possibility since the sample volumes per square centimeter of filter in the work of Van Vaeck et al. (5) (-0.8 m3/cm2)were similar to those used in the present study (-0.8 to - 4 m3/cm2). Corrections for gas adsorption on filters were not made in two of our previous papers (20,21)concerning precipitation scavenging of organic compounds, since the potentially compensatory effects of volatilization losses were not measured. For the sake of completeness, corrections for gas adsorption alone were calculated here. The corrections are fairly small when compared to other uncertainties. Moreover, the general distinctions between the relative importances of gas and particle scavenging made in those papers remain unchanged for all compounds. Since gas adsorption leads to an overestimation of the particulate-phase concentration, it causes an underestimation of particle scavenging ratio ( W,) values. As expected, the alkanes required the largest correction factors. For the mean W, values cited earlier (21),the mean cor-
-
78
Environ. Sci. Technol., Vol. 23, No. 1, 1989
rection factors are as follows: eicosane, 1.6; heneicosane, 1.4; docosane, 1.5. GFF/TMF Comparison. Since sampling artifacts can obviously affect the accuracy of measured cp and cg values, an examination of collection on both GFFs and TMFs was included in this study. Prior investigations of filter artifacts (12,15-17,32) have included examinations of (1)the effects of different types of filters on measured levels of ambient atmospheric PAHs, as well as (2)the recoveries of PAHs spiked onto different filter surfaces. In studies involving the collection of ambient PAHs on GFFs, quartz fiber filters (QFFs), and TMFs, the amounts of PAHs found on TMFs have often exceeded the amounts found on GFFs or QFFs, leading to GFF/TMF and/or QFF/ TMF ratios that are less than 1.0 (16,17). The mechanism usually proposed as the cause is catalytic degradation on the glass or quartz surface. This degradation might take place at the interface between a collected particle and the filter surface, or simply on the filter surface alone. The latter might occur following volatilization from a particle than readsorption onto the filter surface. When compounds are spiked directly onto blank filters, reactions at particle/filter interfaces will clearly not be involved. However, in studies involving the spiking of compounds onto filters also holding precollected particulate matter, both filter and particle/fiIter catalysis could be occurring. In addition to filter-catalyzed degradation, a second possible mechanism for higher TMF levels is an enhanced adsorption of gas-phase compounds on the TMF surface (16). Also, using radioactive benzo[a]pyrene (BaP), Lee et al. (15) have obtained evidence that suggests that either GFF material itself or species sorbed onto it during sampling can lead to BaP degradation during the filter extraction step. Finally, although it has not yet been suggested, it seems possible that the different profiles of the pressure drops through GFFs and TMFs could cause different volatilization losses of the "blow-off" type discussed by Van Vaeck et al. (5). That is, although the overall pressure drop across a TMF will tend to be greater than that across a GFF, the type of bed filtration that takes place with a GFF could lead to comparatively more exposure of collected particles to reduced pressure than would occur with a TMF. With the latter, the bulk of the pressure drop occurs in the interior of the porous membrane where there are few collected particles. The results of the GFF/TMF comparison in the 1985 Portland sampling revealed few significant differences in the concentrations of most PAHs and oxo-PAHs. The mean GFF/TMF ratios and the significance levels for their deviations from 1.0 with a two-tailed test are presented in Table 111. Only phenanthrene was collected in significantly (295% confidence level) lower quantities on the GFFs, and only 7,12-benz[a]anthracenedioneand acenaphthylene were significantly higher on the GFFs. The mean GFF/TMF ratios for the remaining compounds fell between 0.58 and 1.16. These results differ somewhat from those of Grosjean (16),who obtained average GFF/TMF ratios of 0.25-0.77 for some of the same compounds in Los Angeles, but are comparable to those of Fitz et al. (In,who obtained ratios of 0.48-1.03 in the same general area (El Monte). In agreement with our work, it was for phenanthrene in one sampling event that Fitz et al. (17) obtained one of the lowest GFF/TFF ratios (0.58). It should be pointed out here that for compounds subject to deviations from 1.0 in the GFF/TMF ratio, there will very likely be no single, typical GFF/TMF ratio. That is, variabilities due to differences in the sampling conditions during different studies are to be expected.
Table IV. Mean Gas- and Particulate-Phase Concentrations (*lo) and Resulting 4 Values in Portland, OR, during February and April 1984 and February and April 1985"
Table 111. Mean GFF/TMF Ratios ( f l u ) for PAH and Oxo-PAHCompounds for Sampling in Portland, OR, during February and March 1985 com Pound
GFF/TMF ratio
pn
acenaphthylene dibenzofuran fluorene phenanthrene methylphenanthrenes fluoranthene pyrene benzota]fluorene benzo[ blfluorene benztalanthracene chrysene benzo [b+j+k] fluoranthene benzo[e]pyrene benzo[a]pyrene perylene indeno[1,2,3-cd]pyrene dibenz[a,c]- dibenz[a,h]anthracene benzo[ghi]perylene coronene 9-fluorenone 9,10-anthracenedione 7-benz[de]anthracenone 7,12-benz[a]anthracenedione
1.43 & 0.27 0.83 A 0.76 0.61 f 0.59 0.55 f 0.24 0.58 & 0.35 0.97 it 0.28 1.06 f 0.24 0.90 f 0.24 0.93 f 0.23 1.06 f 0.24 1.11 f 0.28 1.10 f 0.24 1.12 f 0.23 0.89 f 0.27 1.05 f 0.26 0.99 & 0.21 0.91 0.23 1.05 f 0.20 0.91 k 0.36 0.81 f 0.19 1.16 f 0.19 1.09 f 0.19 1.27 f 0.10
0.05 0.68 0.28 0.03 0.09 0.86 0.70 0.47 0.63 0.66 0.48 0.47 0.39 0.48 0.76 0.90 0.48 0.64 0.65 0.14 0.19 0.40 0.01
+
", probability that GFF/TMF ratio is the same as 1.0. p I 0.05 indicates a significant (95% confidence level) difference between GFF and TMF values. BaP is of particular interest in a GFF/TMF comparison because several studies have been carried out on its reactivity. Pitts et al. (11)found some conversion of BaP to various oxygenated and nitrated compounds when milligram quantities were spiked onto blank filters and exposed to a flow of ambient air. Brorstrom et al. (12) found losses of up to 40% for ambient BaP on GFFs when 1 ppm NOz was added to the airstream. Grosjean et al. (33) found no degradation when BaP was placed onto GFFs and TMFs and exposed to ambient air as well as purified air spiked with low levels of oxidants, i.e., 100 ppb NOz,SOz, or 03. Blank filters as well as filters loaded with ambient, diesel, and fly ash particles were studied by Grosjean (33). Degradation was only observed when HN03 was present. Although no significant differences in the concentrations of BaP measured on the two types of filters are apparent in Table 111, a close examination of the BaP levels in conjunction with the levels of benzo[e]pyrene (BeP), a less reactive isomer (34,351,is warranted. The BaP/BeP ratios were therefore computed for the four events sampled in 1985 in Portland. The mean ratios ( f l u ) were found to be 1.20 f 0.19 and 1.55 f 0.06 for the GFFs and TMFs, respectively. Thus, the ratio was found to be significantly higher (p = 0.05) as well as more consistent on the TMFs, and so it is possible that some slight losses of BaP did occur on the GFFs. Those losses may have been kept low by virtue of the fact that the oxidant levels were also low in Portland during the sampling. The fact that particulate concentrations of most PAHs were largely the same whether measured on GFFs or TMFs indicates that for this investigation, within the uncertainties of the sampling and analytical methods [f15 and f1870,respectively (23)],filter-catalyzed degradation of most PAHs was not a large problem. It also indicates that gas adsorption of PAHs was not appreciably greater on GFFs than on TMFs. Since measurements of total organic carbon on TMFs have been found to be affected in only a small way by gas adsorption (8),the general comparability of the GFF/TMF results supports the conclusion
compound
gae, ns/ ma
32 f 24 19 f 9 6.8 f 5.6 11 f 7 7.0 f 2.5 (7.0 f 2.5) dibenzothiophene 1.8 1.0 phenanthrene 26 f 10 (26 f 10) anthracene 3.4 h 2.2 xanthone 1.5 0.7 7.2 f 3.8 2- + 3-methylphenanthrene (7.2 f 3.8) 1- + 4- + 9-methyl- 5.7 f 2.8 phenanthrene (5.8 f 2.8) 9,lO-anthracene2.5 f 1.0 dione (2.5 f 1.0) eicosane 4.8 f 2.6 (5.4 f 3.0) fluoranthene 7.9 h 3.1 (8.0 f 3.1) heneicosane 2.6 f 1.3 (3.4 f 2.5) pyrene 6.7 f 2.7 (6.8 f 2.7) benzo[a]fluorene 1.6 f 0.8 benzo[b]fluorene 1.5 f 0.7 docosane 1.4 f 0.4 (1.9 f 1.0) butyl benzyl 5.0 f 0.8 phthalate benz [a]0.32 f 0.14 anthracene (0.37 f 0.21) 0.49 f 0.17 chrysene (0.66 f 0.30) 0.067 f 0.029 7-benz[de]anthracenone 0.39 f 0.39 dioctylphthalate benzo[b+j+k]0.11 f 0.12 fluoranthene
acenaphthylene dibenzofuran 1-+ 2-naphthol fluorene 9-fluorenone
*
particulate, ng/ ms
0.021 f 0.011 0.10 f 0.11 0.25 f 0.20 0.067 f 0.076 0.14 f 0.14 (0.12 f 0.14) 0.039 h 0.041 0.28 h 0.26 (0.27 f 0.23) 0.035 f 0.020 0.060 f 0.035 0.19 f 0.13 (0.15 f 0.09) 0.16 f 0.10 (0.12 f 0.08) 0.59 f 0.22 (0.52 f 0.18) 0.88 f 0.63 (0.30 f 0.17) 0.53 f 0.31 (0.42 f 0.25) 1.1 f 1.1
(0.69 f 0.53) 0.62 f 0.37 (0.53 f 0.31) 0.43 f 0.30 0.45 f 0.32 2.9 f 2.6 (2.1 1.8) 4.1 f 3.1
*
1.2 f 0.8 (1.2 f 0.8)
1.5 f 0.9 (1.4 0.9) 1.7 f 1.2
*
*
0.48 0.25 3.6 f 1.9
6 0.0010 f 0.0005 0.0024 f 0.0025 0.055 f 0.054 0.0062 f 0.0075 0.018 f 0.014 (0.016 0.013) 0.023 0.025 0.010 f 0.007 (0.010 f 0.006) 0.009 f 0.004 0.039 f 0.021 0.025 f 0.013 (0.020 f 0.009) 0.027 f 0.013 (0.020 f 0.011) 0.19 f 0.04 (0.17 0.04) 0.15 f 0.07 (0.059 0.025) 0.061 0.027 (0.049 f 0.023) 0.33 f 0.14 (0.19 f 0.09) 0.083 f 0.038 (0.071 f 0.032) 0.19 f 0.09 0.22 f 0.10 0.47 0.21 (0.31 f 0.17) 0.42 0.25
* *
* * *
* *
0.76 f 0.10 (0.73 0.11) 0.72 0.11 (0.65 f 0.12) 0.97 f 0.02
* *
0.56 f 0.30 0.96 f 0.04
a Values in parentheses are corrected for gas absorpItion with backup filter data.
that gas adsorption did not cause major errors in the thermal/optical measurements of total particulate carbon (TPC) on the GFFs. Gas/Particle Distributions. The mean blank-corrected gas and particulate phase concentrations in Portland in 1984-1985 are given in Table IV together with the resulting $I values. The values are based directly on the data from the front filters and the sorbents. For those compounds found on the backup filters, the values were also corrected for gas adsorption with the backup filter data and (6) Acorr = A + 2Fbackup = Ffront - Fbackup
(7)
Equations 4 and 5 incorporate the assumption that gas adsorption occurred to the front and backup filters to an equal extent. The corrected data are given in parentheses in Table IV. The factor of 2 in eq 4 arises from the fact that both the backup filter concentration and an equal amount due to the assumed sorption on the front filter must be added to the gas concentration. Since backup filters were not used in the 1985 samplings, the corrections of the 1985 data utilized the average backup filter percentage for each compound from the 1984 samples. Because contributions of the volatilization artifact were not measured in this study, the corrected values may not necessarily be more correct than the measured values. Environ. Sci. Technol., Vol. 23,No. 1, 1989 79
'"
'
" "
"
"
" "
"
vapor pressures of the solid compounds, the plot is sig-
"4 i
a
J
p /
BbF I
0: c h r L A
4
Slope I O 9 Intercept 6 4 6 r2 093
-1A
log p:
(torr)
Figure 2. Log A l F for PAH compounds vs log p; (Torr) at 7 OC for samples obtained in Portland, OR, in 1984 and 1985. Acy, acenaphthylene; Fi, fluorene; Ph, phenanthrene; An, anthracene; Fln, fluoranthene; Py, pyrene; BaF, benzo[a ]fluorene; BaA, benzo[a 1anthracene: Chr, chrysene. The bars represent f l u for the multiple samplings.
Nevertheless, the former are useful as indicators of the uncertainties in A and Ffiont.It is reassuring that only the alkanes required significant corrections. In agreement with other studies conducted in urban the 4 values for the unsubstituted PAHs areas (3-5,36,37), at the Portland site were CO.1 for three rings,0.05-0.8 for four rings, and 1.0 for five rings. Since the mean temperatures at the Portland and coastal sampling sites (7and 9 "C,respectively) were fairly close, it may be assumed that the compound-dependent p" values were nearly the same for the two sets of samplings. If, in addition, the particulate matter in Portland and at the coast were similar in both size distribution and sorption characteristics, it might be assumed that the compound-dependent values of c and C were also approximately constant between the two sites. For each compound, then, eq 1 and 5 yield OPort/Ocoast = TSPPort/TSPcoast = (1- 4coast)4port/(1 - 4Port)4coast. The values obtained varied between 0.7 (anthracene) and 8.6 (7-benz[de]anthracenone). The mean value was 2.5. Since the mean TSP level in Portland is only -30 pg/m3 (38), Oport/Ocoast and TSP,,/TSP,,, might indeed be only -3. To the extent that (1)artifacts are not problematic, (2) 0 is indeed proportional to TSP, and (3) the values of c and C are constant within a compound class, then for a given sampling for that compound class, both eq 2 and 5 predict that a plot of log A / F vs log p o will be linear with slope +1 and intercept -log c0 = log C/TSP. Different samplings carried out at similar temperatures but different 6 (TSP) values will lead to different, but parallel lines. Therefore, under these assumptions, when mean data for those multiple samplings are averaged and plotted, then the slope will still be +LO and the intercept will be the mean value of -log c0, which will in turn be equal to the mean value of log C/TSP. Figure 2 is a plot of log A / F vsolog(vapor pressure for the subcooled liquid) (i.e., vs log p L )for the Portland PAH data. (The p ; values used for Figure 2 were corrected to 7 "C.) In this case as well as those discussed below, the values of log Acorr/Fco,are quite similar to the log A / F values, and so the former are not plotted. The bars on each of the points represents A l q for the data from the multiple samplings. The values of pLused were from the literature (39-41). When the data in Figure 2 are plotted vs the
-
80
Environ. Sci. Technol., Vol. 23, No. 1, 1989
nificantly less regular (23). [The worst outliers in that context are anthracene and benz[a]anthracene (23).] The points in Figure 2 for acenaphthylene and fluorene deviate from the rough linearity defined by the other data. When the latter are subjected to a linear fit, a slope of 1.09 and intercept of 6.46 are obtained (9= 0.93). As expected, the slope is near 1.0. When the data are assumed to have a slope of +LOO, the best fit intercept is 5.99. When the whole data set is subjected to a linear fit, the best fit slope is 0.86 and the intercept is 5.20 (r2 = 0.93). It is useful to examine the intercept value (5.99) found for the less volatile compounds in Figure 2 when the data is assumed to have a slope of +1.00. As discussed above, according to Junge's (2) model, this intercept should equal -log c0. A value for 0 of -4 X lo4 cm2/cm3can be calculated by assuming the particle size distribution reported for Los Angeles (42) and the average Portland TSP level of -30 wg m-3. This gives an average c value of -0.3 for phenanthrene through chrysene at 7 "C. This is somewhat smaller than the c value of 1.3 discussed by Pankow (1) for PAHs at -20 "C. Since increasing the temperature will tend to reduce c [Pankow ( I ) ] ,the difference in temperature cannot account for the difference in c values. However, -4 X lo4 cm2/cm3is likely to be an overestimation for 0 for particulate matter in Portland. Indeed, the specific surface area for Portland particulate matter (ATSP,cm2/wg) is likely to be significantly smaller than for Los Angeles where gas-to-particle reactions can lead to significant numbers of particles in the submicron size range. Although the degree of correlation for the Figure 2 data is the same (r2 = 0.93) regardless of whether or not the points for acenaphthylene and fluorene are included, these two compounds may have been behaving differently than the others. The fact that the correlation line for phenanthrene through chrysene alone gives a slope very close to the theoretical value of 1.0 is further evidence in support of that conclusion. It is not immediately clear why acenaphthylene and fluorene may have behaved differently. Even if a 30% filter adsorption effect is assumed for them (i.e., just below the filter detection limit), their points would still not fall near the best fit line for the other compounds. Also, since volatilization losses for them would have caused the measured A / F values to be too high, such losses would have tended to alleviate any nonlinearity and not be its cause. On the basis of the above, we conclude that the nonlinearity in Figure 2 is substantially real and not due to artifacts. More than one explanation can be invoked for this result. In terms of Junge's (2) model, the values of c for acenaphthylene and fluorene might have been comparatively greater than for the other compounds. Alternatively, portions of all of the PAHs might have been trapped within particles where rapid exchange with the gas phase could not occur. In the case of the lower volatility PAHs, this would have comparatively small effects on the measured A / F values since significant fractions of the nontrapped portions of those compounds would be adsorbed to the particles anyway. This would leave their log A / F values similar to what would be expected for full equilibrium and leave the expected linearity for log A / F vs log pLlargely intact. In the case of the most volatile compounds, however, at full equilibrium, extremely low values of $ would be expected. Therefore, for such compounds, even very small amounts of nonexchangeable material in the particulate phase would significantly diminish their measured log A / F values. As shown by
4
i
, 4
-I
BbF
-I 0
O
I! i
/
i
Chr3
4
109
4 '
Slope 0 9 8 Intercept 616 12 095
C h r ~ B a A
I
i
-1;
-2:-8 I
I
'
Intercept 7 4 2 r 2 095
i
-, -5 -5 ' i -6 -5 -2 -7
-4
l o g p:
-3
- 2 +- E' _
-1
" " " "
log p:
(torr)
Figure 3. Log A l f for PAH compounds vs log p; (Torr) at 9 OC for samples obtained at the Oregon coast in 1985. The bars represent 1Q for the multiple samplings.
*
Pankow (43), only -0.02% of nonlabile material could cause the downward shift from the linear correlation line observed here. This second explanation is not inconsistent with the repartitioning hypothesis identified elsewhere (21) as a possible reason for the relatively larger particle preobserved for the lower cipitation scavenging ratios ( W,,) molecular weight PAHs. That hypothesis involved the supposition that the exchangeable fraction is larger for the lower molecular weight PAHs than it is for the higher molecular weight PAHs, thus allowing them to repartition more extensively onto larger, more easily scavenged particles. Certainly the presence of a relatively larger fraction of exchangeable material is still possible for a given compound if only -0.02% of it is restricted as being nonexchangeable. As a third explanation for the nonlinearity at high log p i , it might be possible that 100% of the compounds are exchangeable, but that the sorptive surfaces of the aerosol are not uniform with respect to the energetics of sorption. In this case, even a relatively small percentage of strong sites could result in the same effects described above for the exchangeable/nonexchangeablecase. For the high energy site explanation to be valid, however, there would have to be sufficient numbers of those sites available for sorption of the less strongly bound compounds, and the extent to which that would occur would depend upon the amounts of the more strongly bound compounds present. Indeed, the latter would tend to displace the former. At the present time, the general result of nonlinearity for PAHs with high p i values remains unconfirmed;othesimilar, linear correlation of log A(TSP)/F vs log pLcarried out by pidleman and Foreman (9) only extended to PAHs with pL values up to about -3. Figure 3 is analogous to Figure 2, but presents data for the Oregon coast. Corrections for gas adsorption were not made since backyp filters were not used during those samplings. The pL were corrected to the mean temperature of 9 OC. The points for acenaphthylene and fluorene again seem to deviate from the linearity exhibited by the other data. When those two compounds are excluded, the remaining points give a best fit line with a slope of 0.98 and an intercept of 6.16 (r2 = 0.95). When the slope is assumed to be 1.00, the best fit intercept is 6.28. When acenaphthylene and fluorene are included, the best fit parameters are 0.76 and 5.04, respectively (r2 = 0.93). Compared to Figure 2 data, the Figure 3 data are shifted
-4 I
'
'
'
- '3 '
'
r
'
-2
"-
(torr)
Figure 4. Log A (TPC)/f for PAH compounds vs log (Torr) at 7 OC for selected samples obtained In Portland, OR in 1984 and 1985. The bars represent * l a for the multiple samplings.
Table V. Average Values (flu) for log A (TPC)/F for PAHs in Portland, OR, for February and April 1984 and February and March 1985" compound acenaphthylene fluorene phenanthrene
log A(TSP)/F
log p i
4.12 f 0.63 3.36 f 0.50 3.03 f 0.34 (3.06 0.35) 2.98 f 0.37 2.20 f 0.33 (2.28 0.34) 2.09 0.34 (2.14 f 0.34) 1.56 f 0.34 1.47 & 0.34 0.44 0.30 (0.53 f 0.32) 0.56 f 0.30 (0.75 f 0.33)
4.60 f 0.63 3.84 f 0.50 3.51 0.34 (3.54 f 0.35) 2.46 f 0.37 2.68 f 0.33 (2.76 0.34) 2.57 f 0.34 (2.62 f 0.34) 2.04 f 0.34 1.95 i 0.34 0.92 f 0.30 (1.01 f 0.32) 1.04 A 0.30 (1.23 i 0 . 3 3 )
-2.10 -3.07 -4.00
*
anthracene fluoranthene
* *
pyrene benzo[a]fluorene benzo[b]fluorene benz[a]anthracene chrysene
log A(TPC)/F
*
*
-4.05 -4.89 -4.80 -5.59 -5.62 -5.96 -6.39
"Log A(TSP)/F values are also given assuming TPC TSP/3 for the events sampled. The overall mean temperature was 7 "C. Gas adsorption corrected values are given in parentheses.
up somewhat as expected based on the value calculated above for the average computed ePort/6'coastratio. (Since the best fit slopes are not 1.00, the corresponding best fit inotercepts are not consistent with an upwards shift at log pL = 0.) This shift toward the gaseous phase has also been found by Marty et al. (44) for PAHs in gas and particulate samples taken during a cruise in the east Atlantic off the coast of southwest Africa, specifically Gabon. Although TSP levels were not measured in this work, TPC values were obtained for several of the samplings in Portland. Since organic compounds might associate most strongly with the carbonaceous portion of the TSP, it seems possible that TPC might be a better choice than TSP for correlating gas/particle distributions with p i . For the sampling conducted in this work, TPC levels in Portland ranged from 7 to 27 pg/m3, or m1l3of the TSP values that have been measured at the same site (38). The quantity log A(TPC)/F is therefore plotted vs log p i (7 "C) for the Portland PAH data in Figure 4. (The data for Figure 4 are presented in Table V.) Since each sampling event was characterized by a single TPC value for all compounds, the data exhibit the same trends present in Figures 2 and 3. The best fit slope and intercept for phenanthrene through chrysene are 1.09 and 7.42, reEnviron. Sci. Technol., Vol. 23, No. 1, 1989
81
spectively (r2 = 0,95). When the slope is assumed to be 1.00, the best fit intercept is 6.95. It is interesting t o note t h a t t h e f l u bars in Figure 4 are not narrower than in Figure 2. T h i s indicates that including TPC in the parameterization did not significantly improve the correlatability of log A / F with log p;. As a result, it is unclear at the present time whether TPC plays a special role in sorption in t h e urban atmosphere. Nevertheless, it can be pointed o u t that t h e phenanthrene through chrysene portion of Figure 4 bears much similarity to plots of A(TSP)/Ffor PAHs at 20 "C reported by other workers. If TPC is assumed to have been equal to TSP/3 for the winter/spring events sampled in Portland, t h e n t h e slope 1.00 intercept of 6.95 (7 "C) is equivalent
-
t o a log C value of 7.55 at t h e sampling temperature of 7
"C. This is very close to the value of -7.5 found for eq 4 for PAHs by Pankow (1) for 20 "C. Table V gives the A ( T S P ) / F values for 7 "C computed by assuming TSP = TPC/3. S u m m a r y and Conclusions T h e extent of t h e adsorption of gas-phase PAHs a n d oxo-PAHs on GFFs was variable, but averaged
